Machine Learning for Physical Computing

Week 2 - Classification

Before I tried to create my own teachable machine, I tried to make sure I could get the demo to work. Here is a video of me and my setup testing the ready, set, go sound example.

I chose to use Teachable Machine’s Image Classification to determine whether my sunglass case is open or not.

To further improve on this exercise I would like the output to correlate with the sunglasses. For example, when the sunglasses case is opened the motorized blinds may open to let in the sun. Another example may be to turn on more lights throughout the apartment.

Week 3 - Run a Model

I chose the thumb up and down classifier to work on this week.

For my final project, I am most interested in working with image classifiers to react to certain environment conditions.

Week 4 - Train a Model

Video will be posted soon

Week 5 - Final Project Proposal

In The Room, a Sculpture

A sculpture that changes lights and moves components to react to the number of people in front of it

I want my sculpture to react to the state of a certain environment that can not be easily measured through sensors. Using classification through ml5, a camera will detect how many people are in front of it and modify the sculpture to evoke different feelings. An important constraint is that the sculpture will be placed in a home and therefore its changes need to be subtle and quiet.

  • Alone: Evoke a sense of coziness and restfulness

  • Two people: Evoke a sense of intimacy and bonding

  • More than three people: Evoke a sense of excitedness

*Color scheme and movements will change based on experimentation. Will create a proof of concept for the final project.


Visual reference:

Code reference:

  • ML5’s PoseNet to determine how many people are in-camera. However, this method may slow down the program due to extraneous calculations

  • ML5’s Image Classifier to train the number of people in the room. Worried that it may not be as accurate in different situations with different people

Concept

The sculpture consists of a circular frame and a string-like object to symbolize the connections and interactions of people. I used a conductive thread to connect LED sequins and to produce a more string-like feeling.